Currently we auto-rotate according to the direction calculated from positioning data, which means that we have a trade-off between being too sensitive (e.g. rotating due to lane changes etc.) and being too late to respond to turns (which will look confusing if making a turn very soon after the previous one).
We could probably fix this by calculating if the current position is near the route polyline and what is the angle of the nearest route polyline segment (excluding very short segments?) and using that angle for rotation. If the position is far from the route polyline then we'd fall back to the current system.
Currently we auto-rotate according to the direction calculated from positioning data, which means that we have a trade-off between being too sensitive (e.g. rotating due to lane changes etc.) and being too late to respond to turns (which will look confusing if making a turn very soon after the previous one).
We could probably fix this by calculating if the current position is near the route polyline and what is the angle of the nearest route polyline segment (excluding very short segments?) and using that angle for rotation. If the position is far from the route polyline then we'd fall back to the current system.